Data-Driven and Cell-Specific Determination of Nuclei-Associated Actin Structure.

Area of Science:

• Cell Biology
• Biophysics
• Computational Biology

Background:

• Quantitative analysis of filamentous actin (F-actin) is difficult due to its complex, interconnected structure.
• Existing methods for F-actin measurement lack reproducibility, hindering accurate mechanobiology research.

Purpose of the Study:

• To introduce a novel machine learning-based methodology for precise quantification and reconstruction of nuclei-associated F-actin.
• To investigate the role of F-actin in nucleocytoskeletal connectivity and nuclear architecture in mesenchymal stem cells (MSCs).

Main Methods:

• Development of a Convolutional Neural Network (CNN) for segmenting actin filaments and nuclei from 3D confocal microscopy images.
• Reconstruction of individual actin fibers by connecting contours across image slices.
• Quantification of F-actin organization, nuclear shape, and the impact of Linker of Nucleoskeleton and Cytoskeleton (LINC) complex disruption.

Main Results:

• Accurate and reproducible measurement of F-actin number, length, and volume.
• Demonstration of F-actin disorganization at the nuclear envelope upon LINC complex disruption in MSCs.
• Observed reduction in actin fiber length and volume, correlating with a less elongated nuclear shape.

Conclusions:

• The developed machine learning approach provides a robust tool for F-actin quantification in mechanobiology.
• Disruption of LINC complexes leads to significant alterations in F-actin organization and nuclear morphology.
• This work establishes a pipeline for creating computational models based on quantitative F-actin data.